Use case
Ad and landing teardown
Name the hook, check the CTA and score the ad for a developer audience.
The problem
Creative reviews are opinions. You want a repeatable score.
What your code does with the answer
Rewrite headlines that do not land in `number` or `fear` and have a weak score.
Reading an ad like a rubric
The ad copy is the state; three questions score it: hook (choice: a specific number, a risk if you do nothing, social proof, or other), clear_offer (is the next action obvious from the CTA) and strength (score from weak to strong for a developer audience). The sample — "Cut your LLM bill 80%", "Jev decides in 200ms", "Get API access" — reads hook=number, a clear CTA and a strong score: a good ad by the numbers, which is exactly what you want before the subjective argument starts.
From scores to revisions
The rewrite rule: rework the headline when the hook reads other AND strength is under your bar; keep CTAs that score high on clear_offer, rewrite the ones that do not. Run it over a campaign in one pass — a few hundred tokens an ad, so a month of creative is a few dollars. What the score cannot replace is testing: it kills weak variants before you spend on them.
Pitfalls specific to ad teardown
Scores are rubrics, not verdicts: a strong-scored ad that tests badly needs the criteria adjusted, not the ad defended. Review hook criteria ("number, fear, social proof") as your audience ages — developer advertising patterns move. And keep the "other" bucket honest: an ad whose hook is genuinely something new is a finding worth reading, not noise to route around.
The request shape
This is the full request the playground sends for this preset — the same JSON your server sends to /v1/systemone with a Bearer key.
curl https://jevai.website/api/v1/systemone \
-H "Authorization: Bearer $JEV_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"state": "Headline: \"Cut your LLM bill 80%.\" Body: \"Jev decides in 200ms. Start free.\" CTA: \"Get API access\".",
"questions": {
"hook": {
"type": "choice",
"instructions": "What kind of hook is the headline?",
"criteria": {
"number": "A specific number or saving",
"fear": "A risk if you do nothing",
"social": "Social proof",
"other": "Something else"
}
},
"clear_offer": {
"type": "noul",
"instructions": "Is the next action obvious from the CTA?"
},
"strength": {
"type": "score",
"instructions": "How strong is this ad for a developer API?",
"criteria": [
"Weak",
"Average",
"Strong"
]
}
}
}'FAQ about this use case
- Is "80% savings" legal to claim?
- Different question — the read scores execution, not truthfulness. Claim substantiation stays with your compliance process; if you need it as a typed check, run it over the supporting evidence separately.
- Does the body matter as much as the headline?
- In this preset, strength reads the whole unit. A hooked headline on a meandering body pulls the score down — and the drill-down questions tell you where.
- What about non-English ad copy?
- Send the ad as-is and keep criteria in your scoring language, or write the criteria in the ad's language. Test both settings on past campaign winners and keep whichever agrees with your judgment.